Multi-Class Skin Cancer Classification Using Modified MobileNet Architecture

One of the top three dangerous forms of cancer is skin cancer that can be fatal and is brought on by broken DNA. Cells develop uncontrolled way as a result of this damaged DNA, and they are growing faster. Skin lesion classification is a crucial task in dermatology as it helps in the early diagnosis and treatment of various skin diseases. To save time, effort, and human lives, early detection of skin lesions requires an automated system that can be relied upon for classification. Lower death and morbidity rates from skin cancer may result from early identification. The purpose of this work is to use deep learning with modified MobileNet architecture to classify skin lesions. The 10,015 unique images in the HAM10000 dataset, along with the associated files, were used. The well-known quantitative measurements (such as validation accuracy, sensitivity, precision, F1-score, specificity, loss, and confusion matrix) are used to test the efficiency of the proposed method. The proposed method resulted in 93.2%, 93%, 93%, 94%, 93%, 99%, and 23% respectively for validation accuracy, sensitivity, precision, F1-score, specificity, and loss. The suggested model resulted better in comparison to other transfer learning models. This proposed method can be effectively applied for robust and dependable classification of skin cancer.